Online motion synthesis using sequential Monte Carlo

Abstract

Figure 1: An example of the synthesized animation (downsampled from the original 30 fps). Frame 1: balancing in the user-specified ready stance. Frames 2,3: The character anticipates that the ball would hit it and dodges down. Frame 4: anticipation pose to get enough leg swing momentum. Frames 5,6,7: swinging the leg around and following with the rest of the body to end up again in the ready stance. The ready stance facing direction was not given as a goal. AbstractWe present a Model-Predictive Control (MPC) system for online synthesis of interactive and physically valid character motion. Our system enables a complex (36-DOF) 3D human character model to balance in a given pose, dodge projectiles, and improvise a get up strategy if forced to lose balance, all in a dynamic and unpredictable environment. Such contact-rich, predictive and reactive motions have previously only been generated offline or using a handcrafted state machine or a dataset of reference motions, which our system does not require.For each animation frame, our system generates trajectories of character control parameters for the near future -a few secondsusing Sequential Monte Carlo sampling. Our main technical contribution is a multimodal, tree-based sampler that simultaneously explores multiple different near-term control strategies represented as parameter splines. The strategies represented by each sample are evaluated in parallel using a causal physics engine. The best strategy, as determined by an objective function measuring goal achievement, fluidity of motion, etc., is used as the control signal for the current frame, but maintaining multiple hypotheses is crucial for adapting to dynamically changing environments.

Domain of Attraction Expansion for Physics-Based Character Control

“…Recent work by Tassa et al [2012] and Hämäläinen et al [2014] are examples of MPC methods. The iLQG method of Tassa et al can have the character get up from arbitrary lying positions on the ground.…”

Sequential Monte Carlo instant radiosity

“…We extend the algorithm to a more general case in Chapter 9. Hämäläinen et al [2014] make use of SMC with heuristic sampling to synthesize physically valid character motion in a dynamic environment. At its core, this problem involves tracking local maxima of a sequence of target functions (f 1 , .…”

A Somatosensory Interaction System based on Kinect

“…Action recognition based on template matching often uses the algorithm of dynamic time warping (DTW). Indexing uses lower-bounding functions to prune out the number of times DTW needs to be run for certain tasks such as clustering a set of time series or finding the time series that is most similar to a given time series [6] [7].…”